GitHub/moto-9609/android_kernel_motorola_exynos9610.git
11 years agotg3: Enable support for timesync gpio output
Nithin Sujir [Mon, 29 Jul 2013 20:58:38 +0000 (13:58 -0700)]
tg3: Enable support for timesync gpio output

The PTP_CAPABLE tg3 devices have a gpio output that is toggled when the
free running counter matches a watchdog value. This patch adds support
to set the watchdog and enable this feature.

Since the output is controlled via bits in the EAV_REF_CLCK_CTL
register, we have to read-modify-write it when we stop/resume.

Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotg3: Implement the shutdown handler
Nithin Sujir [Mon, 29 Jul 2013 20:58:37 +0000 (13:58 -0700)]
tg3: Implement the shutdown handler

Also remove the call to tg3_power_down_prepare() in tg3_power_down()
since tg3_close() calls it.

Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotg3: Allow NVRAM programming when interface is down
Nithin Sujir [Mon, 29 Jul 2013 20:58:36 +0000 (13:58 -0700)]
tg3: Allow NVRAM programming when interface is down

Previously, when the interface was brought down, the driver would set
the power state to D3hot. In D3hot, we don't have access to the NVRAM.
This patch removes the call to set the power state to PCI_D3hot in
close. A following patch will implement the shutdown handler to properly
set the D3hot state when the system is going down.

Doing the above means that the TG3_PHYFLG_IS_LOW_POWER should not be
checked to validate access to the NVRAM.

Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotg3: Remove incorrect switch to aux power
Nithin Sujir [Mon, 29 Jul 2013 20:58:35 +0000 (13:58 -0700)]
tg3: Remove incorrect switch to aux power

During probe, the driver is incorrectly switching the power to Vaux on
the 5717 and later devices. At this point, we are in D0 state and
drawing maximum power. We also definitely have Vmain available. It
doesn't make sense to switch to Vaux since it has a lesser maximum power
draw and we might go over the limit. On a new system, we observe that
not all ports are recognized in some of the slots with this call in
place.

Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocnic: Update version to 2.5.17 and copyright year.
Michael Chan [Mon, 29 Jul 2013 02:04:00 +0000 (19:04 -0700)]
cnic: Update version to 2.5.17 and copyright year.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocnic: Add missing error checking for RAMROD_CMD_ID_CLOSE
Eddie Wai [Mon, 29 Jul 2013 02:03:59 +0000 (19:03 -0700)]
cnic: Add missing error checking for RAMROD_CMD_ID_CLOSE

Completion status field should also be checked for non-zero error
condition.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocnic: Update TCP options setup for iSCSI.
Eddie Wai [Mon, 29 Jul 2013 02:03:58 +0000 (19:03 -0700)]
cnic: Update TCP options setup for iSCSI.

Update TCP delayed ACK and timestamp options setup to match latest bnx2x
firmware.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocnic: Reset tcp_flags during cnic_cm_create().
Eddie Wai [Mon, 29 Jul 2013 02:03:57 +0000 (19:03 -0700)]
cnic: Reset tcp_flags during cnic_cm_create().

Without resetting it, the bnx2i driver cannot use different options for
different iSCSI connections.

Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocnic: Simplify cnic_release().
Michael Chan [Mon, 29 Jul 2013 02:03:56 +0000 (19:03 -0700)]
cnic: Simplify cnic_release().

Since unregister_netdevice_notifier() will replay the NETDEV_DOWN and
NETDEV_UNREGISTER_EVENTS, the cnic_dev_list will be cleaned up automatically.
The loop to cleanup the cnic_dev_list can be removed in cnic_release().

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocnic: Simplify netdev events handling.
Michael Chan [Mon, 29 Jul 2013 02:03:55 +0000 (19:03 -0700)]
cnic: Simplify netdev events handling.

After this earlier commit to simplify probing:

    commit 4bd9b0fffb193d2e288f67f81821af32df8d4349
    cnic, bnx2x, bnx2: Simplify cnic probing.

we can now reliably receive netdev events and we can simplify the handling
of these events.  We now remove the logic that tries to handle missed
NETDEV_REGISTER events.

This change will allow cleanup to be simplified in the next patch.  We can
now rely on the play back of netdev events during
unregister_netdevice_notifier() to cleanup the structures.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/mlx4_core: Respond to operation request by firmware
Yevgeny Petrilin [Sun, 28 Jul 2013 15:54:21 +0000 (18:54 +0300)]
net/mlx4_core: Respond to operation request by firmware

This commit adds new firmware command and new firmware event.  The firmware
raises the MLX4_EVENT_TYPE_OP_REQUIRED event in order to signal the driver it
needs to perform an administrative operation throughout the MLX4_CMD_GET_OP_REQ
command. At the moment the supported operation is adding/removing multicast
entries which are used by the firmware for handling NCSI traffic in B0
steering mode.

Also, had to swap the order of mlx4_init_mcg_table() and
mlx4_init_eq_table() to make sure that driver will get events only after
resources are initialized to handle it.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.com>
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/mlx4_en: Fix BlueFlame race
Eugenia Emantayev [Thu, 25 Jul 2013 16:21:23 +0000 (19:21 +0300)]
net/mlx4_en: Fix BlueFlame race

Fix a race between BlueFlame flow and stamping in post send flow.
Example:
SW: Build WQE 0 on the TX buffer, except the ownership bit
SW: Set ownership for WQE 0 on the TX buffer
SW: Ring doorbell for WQE 0
SW: Build WQE 1 on the TX buffer, except the ownership bit
SW: Set ownership for WQE 1 on the TX buffer
HW: Read WQE 0 and then WQE 1, before doorbell was rung/BF was done for WQE 1
HW: Produce CQEs for WQE 0 and WQE 1
SW: Process the CQEs, and stamp WQE 0 and WQE 1 accordingly (on the TX buffer)
SW: Copy WQE 1 from the TX buffer to the BF register - ALREADY STAMPED!
HW: CQE error with index 0xFFFF  - the BF WQE's control segment is STAMPED,
so the BF index is 0xFFFF. Error: Invalid Opcode.
As a result QP enters the error state and no traffic can be sent.

Solution:
When stamping - do not stamp last completed wqe.

Signed-off-by: Eugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopktgen: add needed include file
Stephen Rothwell [Mon, 29 Jul 2013 06:21:26 +0000 (16:21 +1000)]
pktgen: add needed include file

Fixes this on PowerPC (at least):

net/core/pktgen.c: In function 'fill_packet_ipv6':
net/core/pktgen.c:2906:3: error: implicit declaration of function 'csum_ipv6_magic' [-Werror=implicit-function-declaration]
   udph->check = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, udplen, IPPROTO_UDP, 0);
   ^

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Sun, 28 Jul 2013 20:18:49 +0000 (13:18 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
This series contains updates to e100 and e1000e.

The e100 patch from Andy simply updates the netif_printk() to use
%*ph to dump small buffers.

The changes to e1000e include a fix from Dean Nelson to resolve a
issue where a pci_clear_master() was accidentally dropped during a
conflict resolution. Wei Young provides 2 patches, one removes an
assignment of the default ring size because it was a duplicate. The
second changes the packet split receive structure to use
PS_PAGE_BUFFERS macro for the length so that problems won't occur
when the length is changed.

The remaining patches for e1000e are from Bruce Allan, where he
provides a number of fixes and updates for I218.  In addition, a
fix for 82583 which can disappear off the PCIe bus, to resolve this,
disable ASPM L1.  Bruce also provides a fix to a previous commit
(commit e60b22c5b7 e1000e: fix accessing to suspended device) so that
devices are only taken out of runtime power management for those
ethtool operations that must access device registers.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipv4, ipv6: send igmpv3/mld packets with TC_PRIO_CONTROL
Hannes Frederic Sowa [Fri, 26 Jul 2013 15:05:16 +0000 (17:05 +0200)]
ipv4, ipv6: send igmpv3/mld packets with TC_PRIO_CONTROL

v2:
a) Also send ipv4 igmp messages with TC_PRIO_CONTROL

Cc: William Manley <william.manley@youview.com>
Cc: Lukas Tribus <luky-37@hotmail.com>
Acked-by: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoe1000e: fix I217/I218 PHY initialization flow
Bruce Allan [Sat, 29 Jun 2013 07:42:39 +0000 (07:42 +0000)]
e1000e: fix I217/I218 PHY initialization flow

The initialization of the PHY on I217/I218, while similar to 82579, must
also check to see if the MAC and PHY are in the same mode (PCIe vs. SMBus)
otherwise the PHY will be inaccessible by the MAC.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: do not resume device from RPM suspend to read PHY status registers
Bruce Allan [Sat, 29 Jun 2013 07:42:25 +0000 (07:42 +0000)]
e1000e: do not resume device from RPM suspend to read PHY status registers

When the device is runtime suspended (e.g. when there is no link), do not
wake it from D3 to read the PHY status; just set the values to typical
power-on defaults as is done when runtime PM is not enabled and there is no
link.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: enable support for new device IDs
Bruce Allan [Sat, 29 Jun 2013 01:15:16 +0000 (01:15 +0000)]
e1000e: enable support for new device IDs

The device IDs 0x15a0 and 0x15a1 are new SKUs that contain the same MAC as
I217 and same PHY as I218.

The device IDs 0x15a2 and 0x15a3 are the same as existing I218 SKUs.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: ethtool unnecessarily takes device out of RPM suspend
Bruce Allan [Thu, 27 Jun 2013 02:44:44 +0000 (02:44 +0000)]
e1000e: ethtool unnecessarily takes device out of RPM suspend

A previous patch (commit e60b22c5b7 e1000e: fix accessing to suspended
device) added .begin and .complete ethtool driver callbacks so that the
device was resumed from Runtime Power Management (RPM) suspend state for
all ethtool operations.  This is overkill for operations which do not need
to access any registers in the device.  This patch makes it so that the
device is taken out of RPM suspend only for those ethtool operations that
must access device registers.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: Tx hang on I218 when linked at 100Half and slow response at 10Mbps
Bruce Allan [Fri, 21 Jun 2013 09:07:13 +0000 (09:07 +0000)]
e1000e: Tx hang on I218 when linked at 100Half and slow response at 10Mbps

Tx hang is an unintended consequence of another workaround that is in the
EEPROM for an issue with the firmware at 10Mbps when K1 (a power mode of
the MAC-PHY interconnect) is enabled.  The issue is resolved by setting
appropriate Tx re-transmission timeouts in the PHY and associated K1 entry
times in the MAC to allow enough transmissions to occur without triggering
a Tx hang.  A similar change is needed when linked at 10Mbps to improve
latency.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: low throughput using 4K jumbos on I218
Bruce Allan [Fri, 21 Jun 2013 09:07:07 +0000 (09:07 +0000)]
e1000e: low throughput using 4K jumbos on I218

Alter the packet buffer allocation accordingly.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: iAMT connections drop on driver unload when jumbo frames enabled
Bruce Allan [Fri, 21 Jun 2013 09:07:02 +0000 (09:07 +0000)]
e1000e: iAMT connections drop on driver unload when jumbo frames enabled

The jumbo frame configuration in the MAC/PHY should be reverted on 82579
and newer parts when the interface is brought down (not just when the MTU
is changed back to standard frame size) otherwise iAMT connections (e.g.
SoL, IDE-R) will be dropped and cannot be re-acquired until the MTU is
changed again.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: disable ASPM L1 on 82583
Bruce Allan [Fri, 21 Jun 2013 09:07:18 +0000 (09:07 +0000)]
e1000e: disable ASPM L1 on 82583

The 82583 can disappear off the PCIe bus.  This device is a modified 82574
which had the same problem which was fixed by disabling ASPM L1; disabling
it on 82583 fixes the issue on this device.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: Use marco instead of digit for defining e1000_rx_desc_packet_split
Wei Yang [Sat, 25 May 2013 06:23:45 +0000 (06:23 +0000)]
e1000e: Use marco instead of digit for defining e1000_rx_desc_packet_split

In structure e1000_rx_desc_packet_split, the size of wb.upper.length is
defined by a digit. This may introduce some problem when the length is
changed.

This patch use the macro PS_PAGE_BUFFERS for the definition. And move the
definition to hw.h.

Signed-off-by: Wei Yang <weiyang@linux.vnet.ibm.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: Remove duplicate assignment of default rx/tx ring size
Wei Yang [Mon, 20 May 2013 17:31:09 +0000 (17:31 +0000)]
e1000e: Remove duplicate assignment of default rx/tx ring size

tx_ring/rx_ring size is assigned in function e1000_alloc_queues(), which is
called by e1000_sw_init() in the early stage of e1000_probe().

This patch just remove the duplicate assignment of this default ring size
value.

Signed-off-by: Wei Yang <weiyang@linux.vnet.ibm.com>
Reviewed-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Reviewed-by: Da Yu Qiu <qiudayu@cn.ibm.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe1000e: restore call to pci_clear_master()
Dean Nelson [Thu, 13 Jun 2013 03:55:44 +0000 (03:55 +0000)]
e1000e: restore call to pci_clear_master()

In attempting to resolve a minor merge conflict, commit e5f2ef7ab4690d2e8faa
(Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net) accidentally
dropped a call to pci_clear_master() that was intended to remain in place.

Commit 4e0855dff094b0d56d6b (e1000e: fix pci-device enable-counter balance)
replaced a call to pci_disable_device() by one to pci_clear_master(). And then
commit 66148babe728f3e00e13 (e1000e: fix runtime power management transitions)
deleted a number of lines starting two lines following that call.

This patch restores the call to pci_clear_master() in __e1000_shutdown().

v2: added summary lines (enclosed in parens) following commit IDs

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Acked-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agoe100: dump small buffers via %*ph
Andy Shevchenko [Wed, 29 May 2013 18:40:36 +0000 (18:40 +0000)]
e100: dump small buffers via %*ph

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
11 years agobonding: remove bond_resend_igmp_join_requests read_unlock leftover
nikolay@redhat.com [Sat, 27 Jul 2013 17:10:10 +0000 (19:10 +0200)]
bonding: remove bond_resend_igmp_join_requests read_unlock leftover

After commit 4aa5dee4d9 ("net: convert resend IGMP to notifier event") we
have 1 read_unlock in bond_resend_igmp_join_requests which isn't paired
with a read_lock because it's removed by that commit.

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopktgen: Use ip_send_check() to compute checksum
Thomas Graf [Thu, 25 Jul 2013 12:08:04 +0000 (14:08 +0200)]
pktgen: Use ip_send_check() to compute checksum

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopktgen: Add UDPCSUM flag to support UDP checksums
Thomas Graf [Thu, 25 Jul 2013 16:12:18 +0000 (18:12 +0200)]
pktgen: Add UDPCSUM flag to support UDP checksums

UDP checksums are optional, hence pktgen has been omitting them in
favour of performance. The optional flag UDPCSUM enables UDP
checksumming. If the output device supports hardware checksumming
the skb is prepared and marked CHECKSUM_PARTIAL, otherwise the
checksum is generated in software.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Ben Greear <greearb@candelatech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoVSOCK: Move af_vsock.h and vsock_addr.h to include/net
Asias He [Thu, 25 Jul 2013 09:39:34 +0000 (17:39 +0800)]
VSOCK: Move af_vsock.h and vsock_addr.h to include/net

This is useful for other VSOCK transport implemented outside the
net/vmw_vsock/ directory to use these headers.

Signed-off-by: Asias He <asias@redhat.com>
Acked-by: Andy King <acking@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'minnow/net-next' of git://git.infradead.org/users/dvhart/linux-2.6...
David S. Miller [Sun, 28 Jul 2013 03:22:56 +0000 (20:22 -0700)]
Merge branch 'minnow/net-next' of git://git.infradead.org/users/dvhart/linux-2.6 into minnow

Darren Hart says:

====================
Add support for the MinnowBoard in the pch_gbe driver. This was
originally sent to LKML as part of the MinnowBoard support series. That
is now partially merged and this version of the patch has been isolated
from those changes and is now completely self-contained.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoUSBNET: increase max rx/tx qlen for improving USB3 thoughtput
Ming Lei [Thu, 25 Jul 2013 05:47:54 +0000 (13:47 +0800)]
USBNET: increase max rx/tx qlen for improving USB3 thoughtput

The default RX_QLEN()/TX_QLEN() didn't consider super speed
USB device, so only max 4 URBs are scheduled at the same time
for tx/rx, then USB3 NIC can't perform very well.

With this patch, both rx and tx thoughput are increased more than
100Mbps when doing iperf test on ax88179_178a USB 3.0 NIC.

Signed-off-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoUSBNET: centralize computing of max rx/tx qlen
Ming Lei [Thu, 25 Jul 2013 05:47:53 +0000 (13:47 +0800)]
USBNET: centralize computing of max rx/tx qlen

This patch centralizes computing of max rx/tx qlen, because:

- RX_QLEN()/TX_QLEN() is called in hot path
- computing depends on device's usb speed, now we have ls/fs, hs, ss,
so more checks need to be involved
- in fact, max rx/tx qlen should not only depend on device USB
speed, but also depend on ethernet link speed, so we need to
consider that in future.
- if SG support is done, max tx qlen may need change too

Generally, hard_mtu and rx_urb_size are changed in bind(), reset()
and link_reset() callback, and change mtu network operation, this
patches introduces the API of usbnet_update_max_qlen(), and calls
it in above path.

Signed-off-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotuntap: hardware vlan tx support
Jason Wang [Thu, 25 Jul 2013 05:00:33 +0000 (13:00 +0800)]
tuntap: hardware vlan tx support

Inspired by commit f09e2249c4f5c7c13261ec73f5a7807076af0c8e (macvtap: restore
vlan header on user read). This patch adds hardware vlan tx support for
tuntap. This is done by copying vlan header directly into userspace in
tun_put_user() instead of doing it through __vlan_put_tag() in
dev_hard_start_xmit(). This eliminates one unnecessary memmove() in
vlan_insert_tag() for 802.1ad and 802.1q traffic.

pktgen test shows about 20% improvement for 802.1q traffic:

Before:
  662149pps 317Mb/sec (317831520bps) errors: 0
After:
  801033pps 384Mb/sec (384495840bps) errors: 0

Cc: Basil Gor <basil.gor@gmail.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/sctp: Refactor SCTP skb checksum computation
Joe Stringer [Thu, 25 Jul 2013 01:52:05 +0000 (10:52 +0900)]
net/sctp: Refactor SCTP skb checksum computation

This patch consolidates the SCTP checksum calculation code from various
places to a single new function, sctp_compute_cksum(skb, offset).

Signed-off-by: Joe Stringer <joe@wand.net.nz>
Reviewed-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovirtio-net: put virtio net header inline with data
Michael S. Tsirkin [Thu, 25 Jul 2013 00:50:23 +0000 (10:20 +0930)]
virtio-net: put virtio net header inline with data

For small packets we can simplify xmit processing
by linearizing buffers with the header:
most packets seem to have enough head room
we can use for this purpose.
Since existing hypervisors require that header
is the first s/g element, we need a feature bit
for this.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobond: cleanup netpoll code
stephen hemminger [Wed, 24 Jul 2013 18:53:57 +0000 (11:53 -0700)]
bond: cleanup netpoll code

This started out with fixing a sparse warning, then I realized that
the wrapper function bond_netpoll_info could just be removed
by rolling it into the enable code.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoteam: cleanup netpoll clode
stephen hemminger [Wed, 24 Jul 2013 18:52:44 +0000 (11:52 -0700)]
team: cleanup netpoll clode

This started out with fixing a sparse warning, then I realized that
the wrapper function team_netpoll_info could just be collapsed away
by rolling it into the enable code.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobridge: cleanup netpoll code
stephen hemminger [Wed, 24 Jul 2013 18:51:41 +0000 (11:51 -0700)]
bridge: cleanup netpoll code

This started out with fixing a sparse warning, then I realized that
the wrapper function br_netpoll_info could just be collapsed away
by rolling it into the enable code.

Also, eliminate unnecessary goto's

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: use pre-defined macro in bond_mode_name instead of magic number 0
Wang Sheng-Hui [Wed, 24 Jul 2013 06:53:26 +0000 (14:53 +0800)]
bonding: use pre-defined macro in bond_mode_name instead of magic number 0

We have BOND_MODE_ROUNDROBIN pre-defined as 0, and it's the lowest
mode number.
Use it to check the arg lower bound instead of magic number 0 in
bond_mode_name.

Signed-off-by: Wang Sheng-Hui <shhuiw@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopch_gbe: Add MinnowBoard support
Darren Hart [Sat, 18 May 2013 21:46:00 +0000 (14:46 -0700)]
pch_gbe: Add MinnowBoard support

The MinnowBoard uses an AR803x PHY with the PCH GBE which requires
special handling. Use the MinnowBoard PCI Subsystem ID to detect this
and add a pci_device_id.driver_data structure and functions to handle
platform setup.

The AR803x does not implement the RGMII 2ns TX clock delay in the trace
routing nor via strapping. Add a detection method for the board and the
PHY and enable the TX clock delay via the registers.

This PHY will hibernate without link for 10 seconds. Ensure the PHY is
awake for probe and then disable hibernation. A future improvement would
be to convert pch_gbe to using PHYLIB and making sure we can wake the
PHY at the necessary times rather than permanently disabling it.

Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Peter Waskiewicz <peter.p.waskiewicz.jr@intel.com>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Joe Perches <joe@perches.com>
Cc: netdev@vger.kernel.org
11 years agodrivers/net/ethernet/stmicro/stmmac: don't check resource with devm_ioremap_resource
Wolfram Sang [Tue, 23 Jul 2013 18:01:45 +0000 (20:01 +0200)]
drivers/net/ethernet/stmicro/stmmac: don't check resource with devm_ioremap_resource

devm_ioremap_resource does sanity checks on the given resource. No need to
duplicate this in the driver.

Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopch_gbe: Use PCH_GBE_PHY_REGS_LEN instead of 32
Darren Hart [Sat, 18 May 2013 21:45:55 +0000 (14:45 -0700)]
pch_gbe: Use PCH_GBE_PHY_REGS_LEN instead of 32

Avoid using magic numbers when we have perfectly good defines just lying
around.

Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Peter Waskiewicz <peter.p.waskiewicz.jr@intel.com>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: netdev@vger.kernel.org
11 years agonet: Make devnet_rename_seq static
Thomas Gleixner [Tue, 23 Jul 2013 14:13:17 +0000 (16:13 +0200)]
net: Make devnet_rename_seq static

No users outside net/core/dev.c.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: TCP_NOTSENT_LOWAT socket option
Eric Dumazet [Tue, 23 Jul 2013 03:27:07 +0000 (20:27 -0700)]
tcp: TCP_NOTSENT_LOWAT socket option

Idea of this patch is to add optional limitation of number of
unsent bytes in TCP sockets, to reduce usage of kernel memory.

TCP receiver might announce a big window, and TCP sender autotuning
might allow a large amount of bytes in write queue, but this has little
performance impact if a large part of this buffering is wasted :

Write queue needs to be large only to deal with large BDP, not
necessarily to cope with scheduling delays (incoming ACKS make room
for the application to queue more bytes)

For most workloads, using a value of 128 KB or less is OK to give
applications enough time to react to POLLOUT events in time
(or being awaken in a blocking sendmsg())

This patch adds two ways to set the limit :

1) Per socket option TCP_NOTSENT_LOWAT

2) A sysctl (/proc/sys/net/ipv4/tcp_notsent_lowat) for sockets
not using TCP_NOTSENT_LOWAT socket option (or setting a zero value)
Default value being UINT_MAX (0xFFFFFFFF), meaning this has no effect.

This changes poll()/select()/epoll() to report POLLOUT
only if number of unsent bytes is below tp->nosent_lowat

Note this might increase number of sendmsg()/sendfile() calls
when using non blocking sockets,
and increase number of context switches for blocking sockets.

Note this is not related to SO_SNDLOWAT (as SO_SNDLOWAT is
defined as :
 Specify the minimum number of bytes in the buffer until
 the socket layer will pass the data to the protocol)

Tested:

netperf sessions, and watching /proc/net/protocols "memory" column for TCP

With 200 concurrent netperf -t TCP_STREAM sessions, amount of kernel memory
used by TCP buffers shrinks by ~55 % (20567 pages instead of 45458)

lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6     1880      2   45458   no     208   yes  ipv6        y  y  y  y  y  y  y  y  y  y  y  y  y  n  y  y  y  y  y
TCP       1696    508   45458   no     208   yes  kernel      y  y  y  y  y  y  y  y  y  y  y  y  y  n  y  y  y  y  y

lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# (super_netperf 200 -t TCP_STREAM -H remote -l 90 &); sleep 60 ; grep TCP /proc/net/protocols
TCPv6     1880      2   20567   no     208   yes  ipv6        y  y  y  y  y  y  y  y  y  y  y  y  y  n  y  y  y  y  y
TCP       1696    508   20567   no     208   yes  kernel      y  y  y  y  y  y  y  y  y  y  y  y  y  n  y  y  y  y  y

Using 128KB has no bad effect on the throughput or cpu usage
of a single flow, although there is an increase of context switches.

A bonus is that we hold socket lock for a shorter amount
of time and should improve latencies of ACK processing.

lpq83:~# echo -1 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local       Remote      Local  Elapsed Throughput Throughput  Local Local  Remote Remote Local   Remote  Service
Send Socket Recv Socket Send   Time               Units       CPU   CPU    CPU    CPU    Service Service Demand
Size        Size        Size   (sec)                          Util  Util   Util   Util   Demand  Demand  Units
Final       Final                                             %     Method %      Method
1651584     6291456     16384  20.00   17447.90   10^6bits/s  3.13  S      -1.00  U      0.353   -1.000  usec/KB

 Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':

           412,514 context-switches

     200.034645535 seconds time elapsed

lpq83:~# echo 131072 >/proc/sys/net/ipv4/tcp_notsent_lowat
lpq83:~# perf stat -e context-switches ./netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3
OMNI Send TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.7.84 () port 0 AF_INET : +/-2.500% @ 99% conf.
Local       Remote      Local  Elapsed Throughput Throughput  Local Local  Remote Remote Local   Remote  Service
Send Socket Recv Socket Send   Time               Units       CPU   CPU    CPU    CPU    Service Service Demand
Size        Size        Size   (sec)                          Util  Util   Util   Util   Demand  Demand  Units
Final       Final                                             %     Method %      Method
1593240     6291456     16384  20.00   17321.16   10^6bits/s  3.35  S      -1.00  U      0.381   -1.000  usec/KB

 Performance counter stats for './netperf -H 7.7.7.84 -t omni -l 20 -c -i10,3':

         2,675,818 context-switches

     200.029651391 seconds time elapsed

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-By: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: add sk_stream_is_writeable() helper
Eric Dumazet [Tue, 23 Jul 2013 03:26:31 +0000 (20:26 -0700)]
net: add sk_stream_is_writeable() helper

Several call sites use the hardcoded following condition :

sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)

Lets use a helper because TCP_NOTSENT_LOWAT support will change this
condition for TCP sockets.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: sctp: trivial: add uapi/linux/sctp.h into maintainers
Daniel Borkmann [Tue, 23 Jul 2013 12:51:48 +0000 (14:51 +0200)]
net: sctp: trivial: add uapi/linux/sctp.h into maintainers

After this file has moved to the uapi section, we also need to update
this in the maintainers file.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: sctp: trivial: update mailing list address
Daniel Borkmann [Tue, 23 Jul 2013 12:51:47 +0000 (14:51 +0200)]
net: sctp: trivial: update mailing list address

The SCTP mailing list address to send patches or questions
to is linux-sctp@vger.kernel.org and not
lksctp-developers@lists.sourceforge.net anymore. Therefore,
update all occurences.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agodrivers: net: cpsw: add support to show hw stats via ethtool
Mugunthan V N [Tue, 23 Jul 2013 10:08:17 +0000 (15:38 +0530)]
drivers: net: cpsw: add support to show hw stats via ethtool

Add support to show CPSW hardware statistics to user via ethtool
so user can find if there were any error reported by hardware or
the system is over loaded duing high data rate transfer.

Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: Fixed up a error "do not initialise statics to 0 or NULL" in bond_main.c
dingtianhong [Tue, 23 Jul 2013 07:25:47 +0000 (15:25 +0800)]
bonding: Fixed up a error "do not initialise statics to 0 or NULL" in bond_main.c

The error is found by the checkpatch.pl tools.

Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: add rtnl protection for bonding_store_fail_over_mac
dingtianhong [Tue, 23 Jul 2013 07:25:39 +0000 (15:25 +0800)]
bonding: add rtnl protection for bonding_store_fail_over_mac

We need rtnl protection while reading slave_cnt and updating
the .fail_over_mac, and it also follows the logic "don't change
anything slave-related without rtnl". :)

Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: bond_sysfs.c checkpatch cleanup
dingtianhong [Tue, 23 Jul 2013 07:25:32 +0000 (15:25 +0800)]
bonding: bond_sysfs.c checkpatch cleanup

net/bonding/bond_sysfs.c:1302: ERROR: else should follow close brace '}'
net/bonding/bond_sysfs.c:1314: ERROR: else should follow close brace '}'

Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: don't call slave_xxx_netpoll under spinlocks
dingtianhong [Tue, 23 Jul 2013 07:25:27 +0000 (15:25 +0800)]
bonding: don't call slave_xxx_netpoll under spinlocks

The slave_xxx_netpoll will call synchronize_rcu_bh(),
so the function may schedule and sleep, it should't be
called under spinlocks.

bond_netpoll_setup() and bond_netpoll_cleanup() are always
protected by rtnl lock, it is no need to take the read lock,
as the slave list couldn't be changed outside rtnl lock.

Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agodrivers/net: enic: Move ethtool code to a separate file
Neel Patel [Mon, 22 Jul 2013 16:59:18 +0000 (09:59 -0700)]
drivers/net: enic: Move ethtool code to a separate file

This patch moves all enic ethtool hooks from enic_main.c to a new file
enic_ethtool.c

Signed-off-by: Neel Patel <neepatel@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Nishank Trivedi <nistrive@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: trans_rdma: remove unused function
Andi Shyti [Mon, 22 Jul 2013 12:59:16 +0000 (14:59 +0200)]
net: trans_rdma: remove unused function

This patch gets rid of the following warning:

net/9p/trans_rdma.c:594:12: warning: â€˜rdma_cancelled’ defined but not used [-Wunused-function]
 static int rdma_cancelled(struct p9_client *client, struct p9_req_t *req)

The rdma_cancelled function is not called anywhere in the kernel

Signed-off-by: Andi Shyti <andi@etezian.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'be2net'
David S. Miller [Wed, 24 Jul 2013 22:41:59 +0000 (15:41 -0700)]
Merge branch 'be2net'

Sathya Perla says:

====================
The following patches are mostly for providing MAC
filtering ability for VFs. Pls apply. Thanks!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobe2net: delete primary MAC address while unloading
Sathya Perla [Tue, 23 Jul 2013 09:55:04 +0000 (15:25 +0530)]
be2net: delete primary MAC address while unloading

Currently the UC-list is being deleted from the HW MAC table, but the primary
MAC is not.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobe2net: use SET/GET_MAC_LIST for SH-R
Sathya Perla [Tue, 23 Jul 2013 09:55:03 +0000 (15:25 +0530)]
be2net: use SET/GET_MAC_LIST for SH-R

On SH-R and Lancer-R, GET_MAC_LIST cmd is better supported
(instead of NTWK_MAC_QUERY cmd) to query provisioned MAC addresses.
Similiarly, (on SH-R and Lancer-R) SET_MAC_LIST must be used by the PF to
provision a permanent MAC addresses to the VF.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobe2net: refactor MAC-addr setup code
Sathya Perla [Tue, 23 Jul 2013 09:55:02 +0000 (15:25 +0530)]
be2net: refactor MAC-addr setup code

The code to configure the permanent MAC in be_setup() has become quite
complicated, with different FW cmds being used for BEx, SH-R and Lancer.
Simplify the logic by moving some of this complexity to be_cmds.c. This
makes the code in be_setup() a little more readable.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobe2net: fix pmac_id for BE3 VFs
Sathya Perla [Tue, 23 Jul 2013 09:55:01 +0000 (15:25 +0530)]
be2net: fix pmac_id for BE3 VFs

For BE3 VFs, the permanent MAC is added by its PF. The VF can retrieve its
pmac_id only via the IFACE_CREATE cmd. This is not true for Lancer and SH-R
VFs which get the pmac_id by issuing a ADD_IFACE_MAC cmd. So, use this
hack only for BE3 VFs.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobe2net: allow VFs to program MAC and VLAN filters
Sathya Perla [Tue, 23 Jul 2013 09:55:00 +0000 (15:25 +0530)]
be2net: allow VFs to program MAC and VLAN filters

In the current design VFs were not allowed to program MAC/VLAN filters.
Only the PF driver was allowed to configure/provision MAC and transparent
VLANs to a VF. Change this to support MAC/VLAN filtering on a VF by a VM.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobe2net: fix MAC address modification for VF
Sathya Perla [Tue, 23 Jul 2013 09:54:59 +0000 (15:24 +0530)]
be2net: fix MAC address modification for VF

Currently, the VFs by default don't have the privilege to modify MAC address.
This will change in a subsequent fix wherein VFs will have the ability to
modify MAC/VLAN filters.

Fix be_mac_addr_set() logic to support MAC address modification on a
privileged VF too.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosh_eth: Add support for r8a7790 SoC
Simon Horman [Tue, 23 Jul 2013 01:18:05 +0000 (10:18 +0900)]
sh_eth: Add support for r8a7790 SoC

This is a copy of support for r8a7778/9 with the .rmiimode mode bit
of struct sh_eth_cpu_data set.

Also update R8A7779 to R8A777x.

Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosh_eth: add support for RMIIMODE register
Simon Horman [Tue, 23 Jul 2013 01:18:04 +0000 (10:18 +0900)]
sh_eth: add support for RMIIMODE register

This register is prsent on the r8a7790 SoC.

Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: ipv6 eliminate parameter "int addrlen" in function fib6_add_1
fan.du [Mon, 22 Jul 2013 06:21:09 +0000 (14:21 +0800)]
net: ipv6 eliminate parameter "int addrlen" in function fib6_add_1

The "int addrlen" in fib6_add_1 is rebundant, as we can get it from
parameter "struct in6_addr *addr" once we modified its type.
And also fix some coding style issues in fib6_add_1

Signed-off-by: Fan Du <fan.du@windriver.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet ipv6: Remove rebundant rt6i_nsiblings initialization
fan.du [Mon, 22 Jul 2013 06:20:12 +0000 (14:20 +0800)]
net ipv6: Remove rebundant rt6i_nsiblings initialization

Seting rt->rt6i_nsiblings to zero is rebundant, because above memset
zeroed the rest of rt excluding the first dst memember.

Signed-off-by: Fan Du <fan.du@windriver.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovti: switch to new ip tunnel code
Amerigo Wang [Sun, 21 Jul 2013 02:46:25 +0000 (10:46 +0800)]
vti: switch to new ip tunnel code

GRE tunnel and IPIP tunnel already switched to the new
ip tunnel code, VTI tunnel can use it too.

Cc: Pravin B Shelar <pshelar@nicira.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Saurabh Mohan <saurabh.mohan@vyatta.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoip6mr: change the prototype of ip6_mr_forward().
Rami Rosen [Sun, 21 Jul 2013 00:00:31 +0000 (03:00 +0300)]
ip6mr: change the prototype of ip6_mr_forward().

This patch changes the prototpye of the ip6_mr_forward() method to return void
instead of int.

The ip6_mr_forward() method always returns 0; moreover, the return value of this
method is not checked anywhere.

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipmr: change the prototype of ip_mr_forward().
Rami Rosen [Sat, 20 Jul 2013 12:09:28 +0000 (15:09 +0300)]
ipmr: change the prototype of ip_mr_forward().

This patch changes the prototpye of the ip_mr_forward() method to return void
instead of int.

The ip_mr_forward() method always returns 0; moreover, the return value of this
method is not checked anywhere.

Signed-off-by: Rami Rosen <ramirose@gmail.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'team' ("add support for peer notifications and igmp rejoins for team")
David S. Miller [Tue, 23 Jul 2013 23:53:03 +0000 (16:53 -0700)]
Merge branch 'team' ("add support for peer notifications and igmp rejoins for team")

Jiri Pirko says:

====================
The middle patch adjusts core infrastructure so the bonding code can be
generalized and reused by team.

v1->v2: using msecs_to_jiffies() as suggested by Eric

Jiri Pirko (3):
  team: add peer notification
  net: convert resend IGMP to notifier event
  team: add support for sending multicast rejoins
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoteam: add support for sending multicast rejoins
Jiri Pirko [Sat, 20 Jul 2013 10:13:54 +0000 (12:13 +0200)]
team: add support for sending multicast rejoins

Similar to what is implemented in bonding. User is able to ask team
driver to send IGMP rejoins in case port is enabled or disabled. Using
previously introduced netdev notifier.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: convert resend IGMP to notifier event
Jiri Pirko [Sat, 20 Jul 2013 10:13:53 +0000 (12:13 +0200)]
net: convert resend IGMP to notifier event

Until now, bond_resend_igmp_join_requests() looks for vlans attached to
bonding device, bridge where bonding act as port manually. It does not
care of other scenarios, like stacked bonds or team device above. Make
this more generic and use netdev notifier to propagate the event to
upper devices and to actually call ip_mc_rejoin_groups().

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoteam: add peer notification
Jiri Pirko [Sat, 20 Jul 2013 10:13:52 +0000 (12:13 +0200)]
team: add peer notification

When port is enabled or disabled, allow to notify peers by unsolicitated
NAs or gratuitous ARPs. Disabled by default.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agomacvlan fdb replace support
Thomas Richter [Fri, 19 Jul 2013 15:20:08 +0000 (17:20 +0200)]
macvlan fdb replace support

Add support for iproute2 command 'bridge fdb replace ...'.
The rtnletlink call back function ndo_fdb_add will be called
with the NLM_F_REPLACE flag set.
Simply return -EOPNOTSUP.

Resubmitted because net-next was closed last week.

Signed-off-by: Thomas Richter <tmricht@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovxlan fdb replace an existing entry
Thomas Richter [Fri, 19 Jul 2013 15:20:07 +0000 (17:20 +0200)]
vxlan fdb replace an existing entry

Add support to replace an existing entry found in the
vxlan fdb database. The entry in question is identified
by its unicast mac address and the destination information
is changed. If the entry is not found, it is added in the
forwarding database. This is similar to changing an entry
in the neighbour table.

Multicast mac addresses can not be changed with the replace
option.

This is useful for virtual machine migration when the
destination of a target virtual machine changes. The replace
feature can be used instead of delete followed by add.

Resubmitted because net-next was closed last week.

Signed-off-by: Thomas Richter <tmricht@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'tcp'
David S. Miller [Tue, 23 Jul 2013 01:01:09 +0000 (18:01 -0700)]
Merge branch 'tcp'

Yuchung Cheng says:

====================
This patch series improve RTT sampling in three ways:
1. Sample RTT during fast recovery and reordering events.
2. Favor ack-based RTT to timestamps because of broken TS ECR fields
3. Consolidate the RTT measurement logic.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: use RTT from SACK for RTO
Yuchung Cheng [Mon, 22 Jul 2013 23:20:48 +0000 (16:20 -0700)]
tcp: use RTT from SACK for RTO

If RTT is not available because Karn's check has failed or no
new packet is acked, use the RTT measured from SACK to estimate
the RTO. The sender can continue to estimate the RTO during loss
recovery or reordering event upon receiving non-partial ACKs.

This also changes when the RTO is re-armed. Previously it is
only re-armed when some data is cummulatively acknowledged (i.e.,
SND.UNA advances), but now it is re-armed whenever RTT estimator
is updated. This feature is particularly useful to reduce spurious
timeout for buffer bloat including cellular carriers [1], and
RTT estimation on reordering events.

[1] "An In-depth Study of LTE: Effect of Network Protocol and
 Application Behavior on Performance", In Proc. of SIGCOMM 2013

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: measure RTT from new SACK
Yuchung Cheng [Mon, 22 Jul 2013 23:20:47 +0000 (16:20 -0700)]
tcp: measure RTT from new SACK

Take RTT sample if an ACK selectively acks some sequences that
have never been retransmitted. The Karn's algorithm does not apply
even if that ACK (s)acks other retransmitted sequences, because it
must been generated by an original but perhaps out-of-order packet.
There is no ambiguity. In case when multiple blocks are newly
sacked because of ACK losses the earliest block is used to
measure RTT, similar to cummulative ACKs.

Such RTT samples allow the sender to estimate the RTO during loss
recovery and packet reordering events. It is still useful even with
TCP timestamps. That's because during these events the SND.UNA may
not advance preventing RTT samples from TS ECR (thus the FLAG_ACKED
check before calling tcp_ack_update_rtt()).  Therefore this new
RTT source is complementary to existing ACK and TS RTT mechanisms.

This patch does not update the RTO. It is done in the next patch.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: prefer packet timing to TS-ECR for RTT
Yuchung Cheng [Mon, 22 Jul 2013 23:20:46 +0000 (16:20 -0700)]
tcp: prefer packet timing to TS-ECR for RTT

Prefer packet timings to TS-ecr for RTT measurements when both
sources are available. That's because broken middle-boxes and remote
peer can return packets with corrupted TS ECR fields. Similarly most
congestion controls that require RTT signals favor timing-based
sources as well. Also check for bad TS ECR values to avoid RTT
blow-ups. It has happened on production Web servers.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: consolidate SYNACK RTT sampling
Yuchung Cheng [Mon, 22 Jul 2013 23:20:45 +0000 (16:20 -0700)]
tcp: consolidate SYNACK RTT sampling

The first patch consolidates SYNACK and other RTT measurement to use a
central function tcp_ack_update_rtt(). A (small) bonus is now SYNACK
RTT measurement happens after PAWS check, potentially reducing the
impact of RTO seeding on bad TCP timestamps values.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'fec'
David S. Miller [Mon, 22 Jul 2013 23:50:38 +0000 (16:50 -0700)]
Merge branch 'fec'

Fabio Estevam says:

====================
This series improves clock handling in the driver by not enabling/disabling
the optional ptp and enet_out clocks unconditionally, check for the return value
of clk_prepare_enable and also handle clk_ptp in suspend/resume.

Remove an unneeded check in platform_get_resource() and also use
devm_request_irq() that can help to simplify the code.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Frank Li <lznuaa@gmail.com>
11 years agofec: Use devm_request_irq()
Fabio Estevam [Sun, 21 Jul 2013 16:25:04 +0000 (13:25 -0300)]
fec: Use devm_request_irq()

Using devm_request_irq() can make the code smaller and cleaner.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofec: Remove unneeded check in platform_get_resource()
Fabio Estevam [Sun, 21 Jul 2013 16:25:03 +0000 (13:25 -0300)]
fec: Remove unneeded check in platform_get_resource()

As devm_ioremap_resource() is used, there is no need to explicitely check the
return value from platform_get_resource(), as this is something that
devm_ioremap_resource() takes care by itself.

Also, place platform_get_resource() prior to devm_ioremap_resource() for
better code readability.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofec: Check the return value from clk_prepare_enable()
Fabio Estevam [Sun, 21 Jul 2013 16:25:02 +0000 (13:25 -0300)]
fec: Check the return value from clk_prepare_enable()

clk_prepare_enable() may fail, so let's check its return value and propagate it
in the case of error.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofec: Enable/disable clk_ptp in suspend/resume
Fabio Estevam [Sun, 21 Jul 2013 16:25:01 +0000 (13:25 -0300)]
fec: Enable/disable clk_ptp in suspend/resume

clk_ptp should also be enabled in fec_resume() and disabled in fec_suspend().

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofec: Fix the order for enabling/disabling the clocks
Fabio Estevam [Sun, 21 Jul 2013 16:25:00 +0000 (13:25 -0300)]
fec: Fix the order for enabling/disabling the clocks

On fec_probe the clocks are enabled in the following order:

clk_ahb -> clk_ipg -> clk_enet_out -> clk_ptp

, so in the error and remove paths we should disabled them in the opposite
order.

Also fix the order in the suspend/resume functions.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofec: Do not enable/disable optional clocks unconditionally
Fabio Estevam [Sun, 21 Jul 2013 16:24:59 +0000 (13:24 -0300)]
fec: Do not enable/disable optional clocks unconditionally

clk_enet_out and clk_ptp are optional clocks, so we should not enable/disable
them unconditionally.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotun: Support software transmit time stamping.
Richard Cochran [Fri, 19 Jul 2013 17:40:10 +0000 (19:40 +0200)]
tun: Support software transmit time stamping.

This patch adds transmit time stamping to the tun/tap driver. Similar
support already exists for UDP, can, and raw packets.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: Provide a generic socket error queue delivery method for Tx time stamps.
Richard Cochran [Fri, 19 Jul 2013 17:40:09 +0000 (19:40 +0200)]
net: Provide a generic socket error queue delivery method for Tx time stamps.

This patch moves the private error queue delivery function from the
af_packet code to the core socket method. In this way, network layers
only needing the error queue for transmit time stamping can share common
code.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/velocity: add poll controller function for velocity nic
Amit Uttamchandani [Fri, 19 Jul 2013 00:45:22 +0000 (17:45 -0700)]
net/velocity: add poll controller function for velocity nic

Add poll controller function for velocity nic.

Signed-off-by: Amit Uttamchandani <auttamchandani@logicube.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agomISDN: replace sum of bitmasks with OR operation.
Alexandru Juncu [Thu, 18 Jul 2013 11:36:48 +0000 (14:36 +0300)]
mISDN: replace sum of bitmasks with OR operation.

Suggested by coccinelle and manually verified.

Signed-off-by: Alexandru Juncu <alexj@rosedu.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/irda: fixed style issues in irttp
Dragos Foianu [Wed, 17 Jul 2013 11:25:38 +0000 (12:25 +0100)]
net/irda: fixed style issues in irttp

Applied error fixes suggested by checpatch.pl

Signed-off-by: Dragos Foianu <dragos.foianu@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 19 Jul 2013 03:08:47 +0000 (20:08 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:
 "A couple interesting SKB fragment handling fixes, plus the usual small
  bits here and there:

   1) Fix 64-bit divide build failure on 32-bit platforms in mlx5, from
      Tim Gardner.

   2) Get rid of a stupid reimplementation on "%*phC" in our sysfs MAC
      address printing helper.

   3) Fix NETIF_F_SG capability advertisement in hyperv driver, if the
      device can't do checksumming offloads then it shouldn't say it can
      do SG either.  From Haiyang Zhang.

   4) bgmac needs to depend on PHYLIB, from Hauke Mehrtens.

   5) Don't leak DMA mappings on mapping failures, from Neil Horman.

   6) We need to reset the transport header of SKBs in ipv4 before we
      attempt to perform early socket demux, just like ipv6 does.  From
      Eric Dumazet.

   7) Add missing locking on vxlan device removal, from Stephen
      Hemminger.

   8) xen-netfront has to make two passes over an SKB to prepare it for
      transfer.  One pass calculates the number of slots needed, the
      second massages the SKB and fills the slots.  Unfortunately, the
      first pass doesn't calculate the number of slots properly so we
      can end up trying to build a MAX_SKB_FRAGS + 1 SKB which doesn't
      work out so well.  Fix from Jan Beulich with help and discussion
      with several others.

   9) Fix a similar problem in tun and macvtap, which have to split up
      scatter-gather elements at PAGE_SIZE boundaries.  Don't do
      zerocopy if it would result in a > MAX_SKB_FRAGS skb.  Fixes from
      Jason Wang.

  10) On receive, once we've decoded the VLAN state completely, clear
      skb->vlan_tci.  Otherwise demuxed tunnels underneath can trigger
      the VLAN code again, corrupting the packet.  Fix from Eric
      Dumazet"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  vlan: fix a race in egress prio management
  vlan: mask vlan prio bits
  macvtap: do not zerocopy if iov needs more pages than MAX_SKB_FRAGS
  tuntap: do not zerocopy if iov needs more pages than MAX_SKB_FRAGS
  pkt_sched: sch_qfq: remove a source of high packet delay/jitter
  xen-netfront: pull on receive skb may need to happen earlier
  vxlan: add necessary locking on device removal
  hyperv: Fix the NETIF_F_SG flag setting in netvsc
  net: Fix sysfs_format_mac() code duplication.
  be2net: Fix to avoid hardware workaround when not needed
  macvtap: do not assume 802.1Q when send vlan packets
  macvtap: fix the missing ret value of TUNSETQUEUE
  ipv4: set transport header earlier
  mlx5 core: Fix __udivdi3 when compiling for 32 bit arches
  bgmac: add dependency to phylib
  net/irda: fixed style issues in irlan_eth
  ethtool: fixed trailing statements in ethtool
  ndisc: bool initializations should use true and false
  atl1e: unmap partially mapped skb on dma error and free skb

11 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 19 Jul 2013 00:39:05 +0000 (17:39 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Peter Anvin:
 "Trying again to get the fixes queue, including the fixed IDT alignment
  patch.

  The UEFI patch is by far the biggest issue at hand: it is currently
  causing quite a few machines to boot.  Which is sad, because the only
  reason they would is because their BIOSes touch memory that has
  already been freed.  The other major issue is that we finally have
  tracked down the root cause of a significant number of machines
  failing to suspend/resume"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86: Make sure IDT is page aligned
  x86, suspend: Handle CPUs which fail to #GP on RDMSR
  x86/platform/ce4100: Add header file for reboot type
  Revert "UEFI: Don't pass boot services regions to SetVirtualAddressMap()"
  efivars: check for EFI_RUNTIME_SERVICES

11 years agoMerge tag 'md-3.11-fixes' of git://neil.brown.name/md
Linus Torvalds [Fri, 19 Jul 2013 00:37:46 +0000 (17:37 -0700)]
Merge tag 'md-3.11-fixes' of git://neil.brown.name/md

Pull md bug fixes from NeilBrown:
 "Sorry boss, back at work now boss.  Here's them nice shiny patches ya
  wanted.  All nicely tagged and justified for -stable and everyfing:

  Three bug fixes for md in 3.10

  3.10 wasn't a good release for md.  The bio changes left a couple of
  bugs, and an md "fix" created another one.

  These three patches appear to fix the issues and have been tagged for
  -stable"

* tag 'md-3.11-fixes' of git://neil.brown.name/md:
  md/raid1: fix bio handling problems in process_checks()
  md: Remove recent change which allows devices to skip recovery.
  md/raid10: fix two problems with RAID10 resync.

11 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Thu, 18 Jul 2013 21:01:08 +0000 (14:01 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "You'll be terribly disappointed in this, I'm not trying to sneak any
  features in or anything, its mostly radeon and intel fixes, a couple
  of ARM driver fixes"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (34 commits)
  drm/radeon/dpm: add debugfs support for RS780/RS880 (v3)
  drm/radeon/dpm/atom: fix broken gcc harder
  drm/radeon/dpm/atom: restructure logic to work around a compiler bug
  drm/radeon/dpm: fix atom vram table parsing
  drm/radeon: fix an endian bug in atom table parsing
  drm/radeon: add a module parameter to disable aspm
  drm/rcar-du: Use the GEM PRIME helpers
  drm/shmobile: Use the GEM PRIME helpers
  uvesafb: Really allow mtrr being 0, as documented and warn()ed
  radeon kms: do not flush uninitialized hotplug work
  drm/radeon/dpm/sumo: handle boost states properly when forcing a perf level
  drm/radeon: align VM PTBs (Page Table Blocks) to 32K
  drm/radeon: allow selection of alignment in the sub-allocator
  drm/radeon: never unpin UVD bo v3
  drm/radeon: fix UVD fence emit
  drm/radeon: add fault decode function for CIK
  drm/radeon: add fault decode function for SI (v2)
  drm/radeon: add fault decode function for cayman/TN (v2)
  drm/radeon: use radeon device for request firmware
  drm/radeon: add missing ttm_eu_backoff_reservation to radeon_bo_list_validate
  ...

11 years agovlan: fix a race in egress prio management
Eric Dumazet [Thu, 18 Jul 2013 16:35:10 +0000 (09:35 -0700)]
vlan: fix a race in egress prio management

egress_priority_map[] hash table updates are protected by rtnl,
and we never remove elements until device is dismantled.

We have to make sure that before inserting an new element in hash table,
all its fields are committed to memory or else another cpu could
find corrupt values and crash.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovlan: mask vlan prio bits
Eric Dumazet [Thu, 18 Jul 2013 14:19:26 +0000 (07:19 -0700)]
vlan: mask vlan prio bits

In commit 48cc32d38a52d0b68f91a171a8d00531edc6a46e
("vlan: don't deliver frames for unknown vlans to protocols")
Florian made sure we set pkt_type to PACKET_OTHERHOST
if the vlan id is set and we could find a vlan device for this
particular id.

But we also have a problem if prio bits are set.

Steinar reported an issue on a router receiving IPv6 frames with a
vlan tag of 4000 (id 0, prio 2), and tunneled into a sit device,
because skb->vlan_tci is set.

Forwarded frame is completely corrupted : We can see (8100:4000)
being inserted in the middle of IPv6 source address :

16:48:00.780413 IP6 2001:16d8:8100:4000:ee1c:0:9d9:bc87 >
9f94:4d95:2001:67c:29f4::: ICMP6, unknown icmp6 type (0), length 64
       0x0000:  0000 0029 8000 c7c3 7103 0001 a0ae e651
       0x0010:  0000 0000 ccce 0b00 0000 0000 1011 1213
       0x0020:  1415 1617 1819 1a1b 1c1d 1e1f 2021 2223
       0x0030:  2425 2627 2829 2a2b 2c2d 2e2f 3031 3233

It seems we are not really ready to properly cope with this right now.

We can probably do better in future kernels :
vlan_get_ingress_priority() should be a netdev property instead of
a per vlan_dev one.

For stable kernels, lets clear vlan_tci to fix the bugs.

Reported-by: Steinar H. Gunderson <sesse@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agomacvtap: do not zerocopy if iov needs more pages than MAX_SKB_FRAGS
Jason Wang [Thu, 18 Jul 2013 02:55:16 +0000 (10:55 +0800)]
macvtap: do not zerocopy if iov needs more pages than MAX_SKB_FRAGS

We try to linearize part of the skb when the number of iov is greater than
MAX_SKB_FRAGS. This is not enough since each single vector may occupy more than
one pages, so zerocopy_sg_fromiovec() may still fail and may break the guest
network.

Solve this problem by calculate the pages needed for iov before trying to do
zerocopy and switch to use copy instead of zerocopy if it needs more than
MAX_SKB_FRAGS.

This is done through introducing a new helper to count the pages for iov, and
call uarg->callback() manually when switching from zerocopy to copy to notify
vhost.

We can do further optimization on top.

This bug were introduced from b92946e2919134ebe2a4083e4302236295ea2a73
(macvtap: zerocopy: validate vectors before building skb).

Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>